Brains by Design

Can machines become truly intelligent? Mathematician John D. Williams of the Rand Corp. told a Tempe, Ariz. meeting of the National Council of Teachers of Mathematics that they can and probably will become more intelligent than man.

Dr. Williams has respect enough for the human brain. It does some things very well, he said, such as recalling almost instantly bits of information that have been in dead storage for years. It is also capable of abstractions as varied as the theory of relativity and the Song of Solomon. It can think about itself, he added, and assign value judgments to what it thinks--such as that its thinking is wonderful. 'Giant electronic computers in existence at present can do none of these things.

Item for Survival. Trouble is, the human brain, said Williams, is badly organized, inaccurate and slow. It is so complicated that to copy it artificially would be practically impossible. But, according to Williams, solving the problem of copying the brain is neither necessary nor desirable, since nature did not design it for intelligence. "The brain of man, like that of other vertebrates, is an item of random design to meet one basic purpose: survival. The fact that it has outthought things like saber-toothed tigers is no evidence that it is particularly apt for abstract thinking."

Creators of intelligent artificial brains, said Williams, should strive for machines that are designed and built specially for abstract thinking. The necessary hardware will soon be available: electronic units, analogous to brain cells, that can be produced by the billion, be made too small to see with a microscope, send 100 million signals per second, never make mistakes and last indefinitely. Computers made of these wonderful gadgets and geared for abstract thought should be able to outthink the brightest human brain.

Williams did not explain how artificial superbrains will be designed or built. He seems to think that the best policy is to encourage existing electronic computers to design their own successors, developing more intelligence at each step. When computer evolution has made its first great advances, human intelligence will withdraw from competition. Humans can do so, said Williams, by defining intelligence so as to exclude the machine achievement.

In a future universe dominated intellectually by brainy machines, Williams sees one faint hope for man: "The science of genetics is young and moving swiftly. We may learn how to design our children on the machines."

Artistic Huncher. Many mathematicians and computer designers agree with Williams, though seldom in as colorful words. But experts on the human brain tend to doubt that machines will ever match it in creative thought, its highest attribute. Says Neurophysiologist Gerhard Werner of Cornell Medical College: "When you talk about deductive reasoning (making judgments on the basis of given facts), I agree that the computers will undoubtedly be faster than a human brain, more efficient, and come up with the right answers. But in the field of inductive reasoning, I do not think machines will ever compete. When scientists formulate their hypotheses, for example, they work on this artistic level. They make their selection of hypotheses by inspired flashes of insight: hunches. I cannot tell you how this works, and I do not know that anyone can. But this seems to be the way that the human brain works, and I do not think a machine will ever be able to do it. A machine might come up by chance with a correct hypothesis, but it would also produce an almost infinite number of hypotheses that are wrong or inadequate."

This file is automatically generated by a robot program, so reader's discretion is required.